Depends on the actuators and wiring you deploy.
The recommended approach is to have at least one light per room with a physical switch attached.
It can be an actuator to have connections to both, light and switch.
You should apply this scheme for anything ‘mission critical’ such as one light per room and some of the HVAC stuff. I for instance use this to override the hot water pumps (the switch input being connected to the heating controller so pumps are still activated when OH is down).
1 Like
Yes, created my first panel… 
… I understand what you’re saying… as I can see after using / having / seeing the panel that this might well be the case (funnily enough).
I read some posts about sluggish panel response and ‘filtering’… must be related to tons of items in REST, as mine works very well so far.
Still thinking and debating the ‘fall-back’ solutions / approach…
There are even portable lamps available which - in case of failure - can be switched on manually. 
Imho the hole fail safe thing is overrated.
Sure - if nothing works anymore it’s super annoying and a huge pain but you can express order replacement parts and have everything up and running again in one to three days.
That’s only something you can answer. What risk are you willing to accept where RISK = Likelihood of failure * Impact. Usually Impact is measured in money but you can come up with some sort of number here, like a scale of 0-100 or the like. When considering the impact, consider the worst case scenario (e.g. the failure occurs and no one can be there to fix it for weeks), as well as average case (e.g failure occurs and someone not you can be there to fix it).
Come up with all the scenarios you can think of, guestimate at the likelihoods, and then rank them based on the calculated risk. The ones at the top are the ones you need to spend the most amount of effort mitigating. Mitigations include:
- redundant systems
- spare hardware
- failing open/failing safe (e.g. if an irrigation valve is ON and OH crashes, the valve shuts itself OFF either immediately or at some future point (e.g. when it was originally scheduled to shut off).
That’s a long time to have a sprinkler running (or not running) without the ability to control it. That’s a long time to not be able to open a garage door, or to have zero control over your lights. That’s a long time to no have cooling in the summer or heating in the winter (in some places either of which could be life threatening).
If your automations not working is merely annoying, than that risk calculation above bares that out. But I happen to know that Max is running everything off of openHAB, including some of the above “mission critical” functions.
And not everywhere in the world has 1-3 day delivery. And not all problems/failures are caused by failed hardware.
1 Like
I have no physical light or rollershutter switches anymore but still don’t have a wall panel - couldn’t find a good one yet. But I have to say, good automation does help, well, and alexa. It’s been like this for a year now.
1 Like
I am with you; as a project manager by trade I can relate to risk management quite well.
Assessing the most prominent risk an average household faces these days, the fridge, freezer and mobile phone come to mind… none of which are usually backed up by replacement hardware or alternatives.
Since light switches and even irrigation (if not used for production) do not fall into these high-risk categories,it supports my approach of not installing any light switches (for example).
What will certainly help is documentation of the systems, and quick guides for trades to understand the picture of how things work. E.g. for lighting: 240V > Power supply (240V/24V) > LED drivers (controlled by OH) > LEDs.
The idea of controllers assuming a default state has been implemented, and I will continue to do so. I use heartbeats via MQTT, which if not received by the respective device trigger either OFF or a default state.
However, I am looking at improving the fault-tolerance of my openHAB system… cluster, back-up, commercial network gear, etc.
And while I titled this post with “wall panel” in mind, which I pretty much solved with HABPanel, the LED wiring business still keeps me awake at night. I asked that question four years ago and still have not progressed, given the complexity of where to put the drivers (central vs close to the LED), wiring (DC vs AC)… The thikning here ATM is to wire everything back to a central space I have literally built for all things technical, network, electrical, OH, NAS, HRV, hot water, heated floor valves, pumps…
Keep the ideas coming; I appreciate it… as this forum is often the only ‘educated’ group I can run my ideas by; thank you. 
I have also been thinking about things like this, though I’m not close to building my own house I am thinking ahead in my home automation.
I have considered ways to power LED lighting throughout the house and had come up with a concept that involved 24VAC wiring going to custom devices in each room. The problem with that plan was control and doubled up cabling. I never tested it before but I have my doubts about how well 100BASE-T ethernet would work in the same cable as noisy 24VAC power.
Recently I came back around to thinking about this problem and how it would be nice to have a medium power PoE device that I could put something like Tasmota on. That way I would be using existing standards which would make the hardware readily available and proven. I did some looking around and there is beta support for the ethernet PHY in the ESP32 that was added a few months ago. Alas the power capacity in the hardware found was the 12.95W class, not going to power many LEDs with that for room lighting.
So I’ve started designing my own device that meets these requirements that so far supports up to 28.8W passive PoE and will be able to support receiving power from an external power adaptor such as a higher power PoE splitter. Would something like that be of interest?
I also want voice hosted locally and have been looking at Mycroft. I have not invested the time to set it up yet, but they do offer a self hosted option for those more technical users willing to set it up.
My first thought is: have an empty pipe from the main power splitter (basement etc) to each of the locations where you plan to have light. That is cheap to install (when building a new house) and you buy yourself some “oh, i change my mind”-options.
I personally forgot that part, I have 240v to the lights without the pipe. SO I will never have the option to power the lights with PoE. But I can always run DC to it and keep the converter hidden, as I do have a cable going straight from the possible light location to basement. And now, I have a mixed setup. Some 24V LEDs, some 12V LEDs, some 5v RGBW LEDS. Some have the converter in the basement, some next to the LEDS. I think I would prefer one big converter in the basement instead of little converters here and there, but then, what difference does it make.
The other option is to have the converter sitting where your light switches would be, to keep this space available if you plan to sell the house to a noob who cannot program their lightning in openhab (meh) 
I hear you… this LED business is doing my head in; given the variations possible; .e.g.
Off-the-self: LED, driver, actuator… with actuator being central or in the ceiling next to the LED.
How many LEDs will one driver serve? LED current will determine driver / LEDs wattage.
DIY: I have developed a 16-channel dimming LED driver (16 x 6W), allowing configs of say 8 tune-able whites, 16 individuals, 5 RGBs.
I also thought of electrical sub-panels per room, to put the driver, power-supply etc in, also in support of short DC wiring… but have given up on that idea (wring all back to a central location).
As for your ideas…
I wouldn’t change a 'standard like PoE to drive LEDs.
I would also separate 24V from networking cables; main reason for the latter: I will be using 1.5mm2, which can be repurposed for 240V should the need/wish arise (hence the separation).
Building in AU is different to mainland Europe… lots of hollow spaces, where drivers can be placed. In the standard AU ‘sh!tbox’ as I refer to it, cables can be run in the ceiling by accessing the roof space, and cables can be dropped into hollow walls. I have rewired my previous house with Ethernet w/o any problem.
However, I am building a passive house, totally unusual in AU; 40 times more energy efficient, compared to the standard AU house. Cabling as an afterthought will be much more difficult if not impossible… mainly due to the house being air-tight; fully covered with a special foil and sealed (yes, sealing every penetration).
Also, each dimmable light requires a driver (which determines the number of ‘converters’ (as you refer to them).
As you describe your house compared to standard-AU, that is how new houses are build here in Germany and with that in Mind, I therefore think those pipes(google pictures for leerrohre) are highly necessary to be flexible after the house is fiinished, especially because you are “trying” new things. At leaste one per wall.
Yes, but that’s like the candles you need on power outages - you never know where they are when you need 'em. Or if you do, you don’t find the matches in the dark.
Yes and no.
Agree when it comes to redundant server setups. Typical for IT guys like ourselves to invest in tricky-complex solutions to mitigate the risk of server HW outage.
But as-typical to apply wrong priorities: other components of your (smart) home are way more likely to fail.
As Rich correctly noted, RISK = Likelihood of failure * Impact.
Having the mission critical parts of your home to work without OH on the one hand side is a real must.
On the other hand side it’s alsolutely sufficient to restrict that to the mission critical components and functionality.
Just too many more things may go wrong (e.g. SW refusing to properly start or initialize again, Internet down so you cannot lookup things(!), …), at the wrong time (on sundays, during power outages or when you’re not a home but your wife is, …).
BTW impact can also be greatly reduced by having simple but ready-to-be-deployed replacements available (but on site and readily configured).
This! I have and documentation
I literally have spare everything.
They also must be somewhat pre-configured/synchronized so it’s just a matter of exchanging components.
And check them regularly. Just found out that my spare maxcube (to control heating) is broken 
Meaning you wouldn’t want to run LEDs off a PoE device? My plan is to use cheap 10/100 Mode B passive power injectors like this one
I want to avoid running 2 cables to each location, but was planning on running PVC conduit as @papaPhiL suggests.
1 Like
We actually have two spare fridges. But the impact is much higher for us than just losing some food. My wife is T1 Diabetic. If we lose the fridge we lose her insulin. That’s another example about how only you can truly answer what poses the greatest risks. You will have unique impacts which can drastically change the calculation.
Have you looked into Last Will and Testament messages? MQTT may have this built in for you.
When an MQTT client connects to the MQTT broker, it registers a LWT message and topic. Then the broker keeps a heatbeat going between the client and itself. When the client stops responding, the broker will publish that LWT message to the LWT topic. Any client can subscribe to that topic and immediately know that the client is offline.
If you publish something like ONLINE as a retained message to the LWT topic when it connects to the broker and have the broker publish OFFLINE as a retained message to the LWT topic when the broker determines that it’s offline, anyone who subscribes to the LWT topic immediately knows whether or not the device is online or offline. And there is no need to implement any heartbeat logic yourself.
I’m no expert but I have read that it doesn’t work at all. You need some distance and/or isolation between data and power.
This. I bet there is more than one reader of this thread who got a flat tire only to discover that the spare was also flat.
1 Like
Regarding using wall panels, this has not really much to do with automation. Your automation will work with or without wall panels.
But there are situations were you would consider if its worth saving wall panels. Think about children, old persons etc… Those who do not want to or cant handle a smartphone the same way you can. If you have guests, friends or family visiting. Dont expect them to see your smarthome as smart as you do.
1 Like
Exactly. In theory the differential signaling should be able to filter out the noise, but I think a high frequency transient load like PWM dimmed LEDs might cause more noise than it can handle. That’s why that one never went past the “I’ve got a crazy idea” stage.
This is a specific topic, as in “disabled” access (disabled for whatever reasons), my argument would be, light switches would fail them too… in particular the joke of a switch common in AU.
That switch is 15mm2 (0.6" square)… my pads on the current panel are 25mm2 (1" square)… and can be modified to suit a difference audience. Also bearing in mind I will eventually use voice to control things.
I still believe some traditional form of ‘actuator’ needs to exist to cater for most… even if it is an abstraction on a touch panel.
I would have think about its feasibility and any code issues…
It certainly is a novel idea
[update] just checked the AU wiring rules, at least in AU it would violate code: “Separate cables are used for LV power and telecommunications.”
I settled on MQTT in a ‘heartbeat’ but certainly need to have a closer look at the protocol and your comment. Yes, I did implement a ‘heartbeat’ but if the functionality exists in the protocol, I will update my code. Thanks for pointing this out.
That seems to say they are outlawing PoE in the building code… I can’t see them doing that. Certainly you have PoE VoIP phones and such over there?
Well code I’m looking at Tasmota because of the ESP32 PHY support. I haven’t used Tasmota since I first started out with OH 3 years ago, but everyone is using it so it must be pretty stable.
Novel idea indeed, I am having to design the hardware myself. Obviously small batch for myself will not be so cheap per unit, but if there is more interest the cost would come down. If you have any suggestions please let me know. Once I have something to show I will share on here.